Matthew Berman describes a system he packaged as the "Ad Spy Kit": seven Claude skills (ad-spy-kit to run it all, resolve, pull, winners, analyze, counter, render). You give it 3 to 5 competitor brands by name, IG handle or Ad Library link. It pulls their live Meta ads through StealAds or Scrape Creators, sorts them by days running and number of versions, and labels winners (60–120+ days with 3+ variations) versus tests and flops. It then pulls hooks and offers word for word from the transcript, on-screen text and caption, with each quote tied to its source ad so you can check it, maps the promises, formats and ignored objections, finds what every brand is saying and which formats are crowded, and outputs a board with a headline, hook, what to show and the offer built around the gap nobody has taken. He frames it as replacing a $149/mo Foreplay plan plus an $8K/mo creative strategist.
← All agent use cases · Creative use cases
Seven Claude skills turn 3–5 competitors' live Meta ads into winners, word-for-word hooks, market gaps and a brief for your next ad
Claude skillsScrape CreatorsStealAdsMeta Ad Library
AdShift lists real agent workflows people have shared publicly. We summarize them and link to the original; check the source before you copy a setup.
More Creative use cases All →
- A Claude Code operator folder plus MCPs: scan competitors' top ads and landing pages, then generate on-brand statics with Higgsfield Nov 9
- Rip a competitor's best-performing ad format, rewrite it with your own customer research, and have Claude generate the UGC video Nov 6
- An agent that writes ad scripts from your winners and losers, diagnoses dying ads from hook and hold rate, and pulls every ad's funnel metrics each Monday Nov 2
- An outbound agent that finds brands running new Meta ads among your LinkedIn connections, makes 4 spec ads for each, and drafts the pitch email Nov 2
Monday brief
One new agent use case every Monday, alongside the platform changes that matter.